The AI Power Crisis Proves Data Centers Must Be Regulated as Public Utilities

The artificial intelligence revolution is colliding with the physical limits of the American electrical grid. On one side of this conflict are technology giants racing to build massive data centers to train the next generation of AI models. On the other side are local communities, environmental advocates, and ratepayers who are suddenly facing the prospect of soaring electricity bills and increased fossil fuel emissions to power that compute. The tension is acute: AI requires an unprecedented scale of energy, but the traditional model of socializing infrastructure costs threatens to penalize everyday citizens for Big Tech's expansion. Resolving this requires a fundamental shift in how we govern digital infrastructure.[6]

The emerging progressive consensus is clear: data centers can no longer be treated as standard commercial real estate. Because they operate as industrial-scale power consumers, they must be regulated with the rigorous oversight traditionally applied to public utilities. This framework argues that when a single facility consumes as much electricity as a small city, its impact on grid stability, water resources, and local utility rates becomes a matter of urgent public interest. Without strict regulatory guardrails, the financial and environmental externalities of the AI boom will be offloaded onto the public.[1][2]

The sheer scale of AI's energy demand illustrates why this regulatory shift is necessary. According to projections from the Department of Energy and the Lawrence Berkeley National Laboratory, data centers consumed roughly 4.4% of total U.S. electricity in 2023. By 2028, driven almost entirely by the continuous, high-density computing required for AI training and inference, that figure is projected to reach between 6.7% and 12%. A single frontier AI data center can demand 50 to 100 megawatts of continuous power. This is not incremental load growth; it is a sudden, massive shock to regional power grids that were not designed for this velocity of expansion.[3]

State Public Utility Commissions (PUCs) are already stepping in to shield residents from the fallout. In states like Ohio and Idaho, regulators are implementing "large-load tariffs." Traditionally, when a utility upgrades the grid to accommodate new demand, the capital costs are socialized across all ratepayers. Large-load tariffs flip this model, legally requiring tech companies to pay the upfront costs for the generation and transmission infrastructure their data centers necessitate. This ensures that residential customers do not see their monthly bills spike simply because a hyperscaler built a facility in their county.[2][5]

Beyond ratepayer protection, environmental advocates are pushing municipal governments to end "by right" zoning approvals for data centers. By requiring special permits, local governments gain the leverage to mandate clean energy usage, strict water conservation standards, and community benefit agreements. If AI developers want access to a community's resources, the progressive argument dictates that they must guarantee their operations will not force local utilities to keep aging, high-emission fossil fuel plants online just to meet the new baseline load.[2]

Some policy advocates are taking the public utility framework a step further, arguing that the AI models and compute infrastructure themselves should be regulated like essential utilities. If frontier AI becomes a necessary input for economic participation, leaving its provision entirely to a few unregulated monopolies risks exclusionary pricing and concentrated power. A public utility model for AI would enforce non-discriminatory access, transparent pricing, and reliability standards, ensuring that the economic surplus generated by AI diffuses broadly across society rather than accumulating solely at the top.[6]

This regulatory push faces fierce opposition from the tech industry and free-market advocates. Critics argue that legacy public utility models are inherently sluggish and ill-equipped to manage the fast-paced innovation of the AI sector. Organizations like the Competitive Enterprise Institute contend that heavy-handed regulation, special tariffs, and permitting moratoriums will artificially bottleneck AI advancement, ultimately driving development to overseas markets with looser standards. They argue the solution to the energy crisis is to deregulate energy markets to spur rapid power plant construction, not to penalize the data centers driving economic growth.[4]

Yet, the physical realities of the grid make a laissez-faire approach increasingly untenable. The era of frictionless, unregulated data center expansion is closing because the electrical infrastructure simply cannot support it without coordinated oversight. By adapting public utility regulations—imposing large-load tariffs, demanding environmental accountability, and protecting residential ratepayers—policymakers are forging a necessary compromise. It is a model that allows the AI industry to grow, but only if it pays its own way and respects the limits of the communities that host it.[1][6]